Raman spectroscopy is a powerful technique that relies on collection of inelastically scattered laser light from a sample. This light exhibits a frequency shift that reflects the energy of specific molecular vibrations within the sample of interest. Hence, it can provide a detailed chemical composition of the sample, i.e. a chemical fingerprint. The technique has wide potential in biomedical science as it may be applied to samples over a wide size range from single cells through to intact tissue.
One of the major challenges of Raman spectroscopy is the inherently weak nature of the signal. In addition, a Raman signal may be obtained from the local environment surrounding the sample, typically making it difficult to discern the molecular signatures of interest. Thus, considerable effort has focused on enhancing the ratio of signal to background noise. By increasing the acquisition time to several minutes, the signal to noise ratio can be improved. However, in some environments, long acquisition times can cause damage due to extended irradiation by the excitation laser and the mechanism required to hold the particles under investigation in the measurement position. These are particular problems when investigating live cells or tissue samples.
Some solutions to the problems with conventional Raman spectroscopy have been proposed. Many of these involve the inclusion of additional material, for example nano-particles, in the samples that are being investigated. However, this is not ideal for the investigation of whole cells as the precise positional control of the foreign particles is difficult. Additionally, the enhancement achieved with the use of foreign particles is confined to the immediate surface of the particles (˜10 nm) making the measurement of the overall Raman signal impossible. One technique that does not require the addition of foreign particles uses wavelength modulation. This is described in the article “Wavelength-Modulation Raman Spectroscopy” by Levin et al, Appl. Phys Letter 33(39), 1 Nov. 1978. This technique increases the sensitivity of a Raman spectroscopic system by modulating the wavelength of the excitation light, and then using this to distinguish the sample's Raman response from background radiation and/or noise. The system described uses a tuneable dye laser and single channel slowly scanning detection. A problem with this is that the scan takes about 50 minutes for the whole spectra. Additionally the method relies on very large, expensive optics and is inappropriate for many practical applications, in particular the investigation of single cells.
One of the most promising areas of application for Raman spectroscopy is in the discrimination between sets of biomedical samples e.g. cancer diagnostics. Here, it is advantageous to have short acquisition times, especially if a live patient rather than a retrieved sample is being studied. It is also important to reduce the impact of fluorescence, as this has a high patient to patient and even cell to cell variability that can heavily reduce the performance of any subsequent diagnostic models. One of the most widely used tools for discriminating between the Raman spectra acquired from sets of biomedical samples is Principal Component Analysis (PCA).
Principal components analysis (PCA) is a statistical technique used to change the representation of a multidimensional data set. A new representation or coordinate system is constructed such that the variance of the data sets is biggest for the first coordinate component of the new representation. This is then called the first principal component. The second biggest variation lies the on the second coordinate of the new representation, and so on. Finally, the data set dimension is reduced by retaining only the first few principal components that account for most of the variance of the original data set. It is these low-order components that often contain the “most important” aspects of the data set. Using PCA to examine Raman spectra from sets of biomedical samples allows combinations of Raman peak fluctuations to be found that can then be used to discriminate between the Raman spectra from the sets of biomedical samples.